Service 04
AI-made video, signed off before it goes out
We produce short videos with generative models: advertising, product demos, recurring social content. AI speeds up the making; it decides nothing. The brief, the edit and the sign-off stay human, and nothing is published without your approval.
Why it works
A video generated in three minutes is worth nothing if nobody decided what it should say. So we start where we start with software: who it speaks to, what it has to make clear, and how we will know it worked. The models come afterwards, as a manufacturing tool — not as an idea.
What they genuinely change is the cost of the variant. Once the format holds, producing ten versions for ten audiences, three aspect ratios and two languages no longer costs a shoot. That is where AI pays for itself, and that is what we industrialise: the series, not the single film.
What it covers
Formats and variants
One format that holds, cut per network and per audience rather than remade each time.
- 9:16 vertical, square and horizontal, each framed on its own
- Alternative openings tested on the same footage
- Short durations calibrated per platform
- Graphics aligned with your brand
- An explicit call to action at the end
Voice, subtitles and languages
Video is often watched with the sound off. Subtitles are not an option.
- A synthetic voice chosen and approved with you
- Burnt-in subtitles, proofread by hand
- Multilingual versions from the same edit
- Voice cloning only with written permission
- Subtitle files delivered separately
Avatars and presenters
A synthetic face commits your brand. The rights are settled before, not after.
- Avatars under a verified commercial licence
- An avatar in your likeness, with written consent
- Lip sync matched to the chosen voice
- Never a third party’s likeness without the same consent
- A synthetic-content disclosure wherever one is required
Post-production
A model’s raw output is not a deliverable. Editing is still the craft.
- Editing, pacing, cuts and transitions
- Grading and visual consistency across shots
- Licensed music and sound, never improvised
- Correction of the artefacts models leave behind
- Export to each platform’s specification
A production chain you keep
The point is not the video we deliver, it is the next series you produce without us.
- Project templates and graphics kits
- A documented library of prompts and settings
- A written procedure: from script to export
- Automated variants once the volume justifies it
- Training for whoever takes it over
What you receive
The videos, in broadcast formats
Masters exported per platform, ready to publish.
The source files
Edit projects, generated footage, graphics — in your hands.
Subtitles and language versions
Separate files, reusable on other edits.
Licences and clearances
Music, avatars, voices: written proof of what you are allowed to broadcast.
The production chain, documented
Prompts, settings, templates, and the procedure to run another series.
A note on the tools used
Which models, for what, and what was sent to them.
The process
Framing and intent
Who the video speaks to, what it has to make clear, and what will count as success. We write the message before choosing a single tool.
Script and pilot format
A short script, then one complete video that becomes the reference. Tone, pacing and graphics are settled on that one — not on ten drafts.
Production and variants
The approved format is cut for audiences, ratios and languages. You review each batch before it continues.
Delivery and handover
Masters, sources, licences and procedure. We show your team how to produce the next batch without us.
Frequently asked questions
Can an AI-generated video be published freely?
Yes, provided three things are under control: the commercial licence of the model used, the rights to the music and the voices, and the consent of anyone whose likeness appears. We deliver nothing without that written trail. We do not, however, give legal advice: on a sensitive use, your own counsel decides.
Does the video have to be labelled as AI-generated?
Platforms and regulation are both moving, and several already require a disclosure for realistic synthetic content. We apply the rule in force at delivery and tell you which one it is. We would rather carry a readable label than argue about intent: content that owns what it is defends itself better than content that is ambiguous.
Can you use a real person’s face or voice?
Yes, with your written permission, and for a defined scope of use: which channels, for how long, and how you can withdraw it. We never use a third party’s likeness — a client, an employee, a public figure — without the same agreement.
What can AI video not do yet?
Models are still weak on character consistency between shots, on text within the image, on hands, and on long scenes. So we frame projects around what holds: short sequences, typography added in the edit, and real filming wherever it remains indispensable. We say so at the framing stage, not at delivery.